modules/route_parser.py

"""Rule-based itinerary route parser for the map tab."""

from __future__ import annotations

import math
import re
import unicodedata

try:
    from .map_builder import (
        DB,
        extract_ordered_locations,
        normalize_location_text,
    )
except ImportError:  # pragma: no cover - supports direct module imports.
    from map_builder import DB, extract_ordered_locations, normalize_location_text


DAY_PREFIX_PATTERN = r"(?:Day|Jour|Giorno|Ngay|Ngày)"
ROUTE_SEPARATOR_PATTERN = re.compile(r"\s*(?:->|>|/|\||;|\s+-\s+|,\s+|\s*&\s+)\s*")
META_CLAUSE_SEPARATOR_PATTERN = re.compile(r"\s*(?:/|\||;|\s*&\s*)\s*")

MODE_KEYWORDS = {
    "plane": {
        "vol", "flight", "flights", "avion", "plane", "fly", "flying", "volo",
        "air", "aerien", "aerienne", "aereo",
    },
    "train": {
        "train", "rail", "railway", "tgv", "treno", "night train", "sleeper train",
    },
    "boat": {
        "boat", "bateau", "cruise", "ferry", "junk", "croisiere", "barca",
        "speedboat", "speed boat", "hydrofoil", "du thuyen",
    },
    "car": {
        "car", "road", "route", "drive", "driving", "driver", "transfer",
        "transfert", "trasferimento", "vehicle", "van", "bus", "coach", "xe",
    },
}

ARRIVAL_WORDS = {"arrival", "arrivee", "arrive", "arrivo"}
DEPARTURE_WORDS = {"departure", "depart", "partenza"}
METADATA_WORDS = ARRIVAL_WORDS | DEPARTURE_WORDS
PENDING_SKIP_WORDS = {
    "airport", "aeroport", "station", "pier", "port", "harbor", "hotel", "resort",
    "homestay", "guide", "chauffeur", "driver", "transfer", "transfert",
    "arrival", "arrivee", "arrive", "departure", "depart", "day", "jour",
}


def _strip_day_prefix(text: str) -> str:
    cleaned = unicodedata.normalize("NFKC", text or "").strip()
    cleaned = re.sub(r"[–—−‐‑‒―]", "-", cleaned)
    return re.sub(rf"^{DAY_PREFIX_PATTERN}\s*\d+[:\s\-–—]*", "", cleaned, flags=re.I).strip()


def _phrase_in_normalized(phrase: str, normalized_text: str) -> bool:
    phrase_norm = normalize_location_text(phrase)
    if not phrase_norm:
        return False
    return bool(re.search(rf"(?<![a-z0-9]){re.escape(phrase_norm)}(?![a-z0-9])", normalized_text))


def _mode_from_text(text: str) -> str | None:
    normalized = normalize_location_text(text)
    if not normalized:
        return None

    matched = []
    for mode, phrases in MODE_KEYWORDS.items():
        if any(_phrase_in_normalized(phrase, normalized) for phrase in phrases):
            matched.append(mode)

    if not matched:
        return None
    for mode in ("plane", "train", "boat", "car"):
        if mode in matched:
            return mode
    return None


def _metadata_kind(text: str) -> str | None:
    normalized = normalize_location_text(text)
    if not normalized:
        return None
    first_token = normalized.split()[0]
    if first_token in ARRIVAL_WORDS:
        return "arrival"
    if first_token in DEPARTURE_WORDS:
        return "departure"
    return None


def _trailing_metadata_kind(text: str) -> str | None:
    normalized = normalize_location_text(text)
    if not normalized:
        return None
    tokens = normalized.split()
    if not tokens:
        return None
    if tokens[-1] in ARRIVAL_WORDS:
        return "arrival"
    if tokens[-1] in DEPARTURE_WORDS:
        return "departure"
    return None


def _remove_metadata_prefix(text: str) -> str:
    return re.sub(
        r"^\s*(?:arrival|arrivee|arriv[eé]e?|arrivo|departure|depart|d[eé]part|partenza)"
        r"\s*(?::|\b)?\s*(?:at|a|à|to|vers|from|de|da)?\s*",
        "",
        text,
        flags=re.I,
    ).strip(" -")


def _resolved_city_list(text: str, include_unknown: bool = True) -> tuple[list[str], list[str]]:
    cities = []
    pending = []
    seen = set()
    for candidate in extract_ordered_locations(text, include_unknown=include_unknown):
        matched = DB.find_city(candidate)
        if matched:
            key = normalize_location_text(matched.name)
            if key not in seen:
                cities.append(matched.name)
                seen.add(key)
        elif include_unknown and _keep_pending_candidate(candidate):
            key = normalize_location_text(candidate)
            if key not in seen:
                pending.append(candidate)
                seen.add(key)
    return cities, pending


def _keep_pending_candidate(candidate: str) -> bool:
    normalized = normalize_location_text(candidate)
    if not normalized or len(normalized) < 3:
        return False
    tokens = normalized.split()
    if not tokens or len(tokens) > 6:
        return False
    if any(token in PENDING_SKIP_WORDS for token in tokens):
        return False
    if any(token in METADATA_WORDS for token in tokens):
        return False
    if _mode_from_text(candidate):
        return False
    return True


def _haversine_km(city_a: str, city_b: str) -> float | None:
    a = DB.find_city(city_a)
    b = DB.find_city(city_b)
    if not a or not b:
        return None
    lat1, lon1 = math.radians(a.lat), math.radians(a.lon)
    lat2, lon2 = math.radians(b.lat), math.radians(b.lon)
    dlat, dlon = lat2 - lat1, lon2 - lon1
    h = math.sin(dlat / 2) ** 2 + math.cos(lat1) * math.cos(lat2) * math.sin(dlon / 2) ** 2
    return 6371 * 2 * math.asin(math.sqrt(h))


def _default_mode(city_a: str, city_b: str) -> str:
    distance = _haversine_km(city_a, city_b)
    if distance is not None and distance > 300:
        return "plane"
    return "car"


def _extract_metadata(line: str) -> tuple[str | None, str | None, str | None]:
    arrival = None
    departure = None
    route_parts = []
    any_metadata = False

    for clause in META_CLAUSE_SEPARATOR_PATTERN.split(line):
        clause = clause.strip()
        if not clause:
            continue

        kind = _metadata_kind(clause) or _trailing_metadata_kind(clause)
        if not kind:
            route_parts.append(clause)
            continue

        cities, _ = _resolved_city_list(clause, include_unknown=False)
        if not cities:
            continue

        any_metadata = True
        if kind == "arrival":
            arrival = arrival or cities[0]
        else:
            departure = cities[-1]

        if len(cities) > 1:
            route_parts.append(_remove_metadata_prefix(clause))

    if not any_metadata:
        return None, None, line

    route_text = " - ".join(part for part in route_parts if part.strip())
    return arrival, departure, route_text or None


def _route_fragments(line: str) -> list[str]:
    line = re.sub(r"\s*&\s*(?=(?:vol|flight|volo)\b)", " - ", line, flags=re.I)
    return [fragment.strip() for fragment in ROUTE_SEPARATOR_PATTERN.split(line) if fragment.strip()]


def _append_leg(legs: list[dict], city_a: str, city_b: str, mode: str | None) -> None:
    if not city_a or not city_b or city_a == city_b:
        return
    legs.append({"a": city_a, "b": city_b, "mode": mode or _default_mode(city_a, city_b)})


def _parse_route_line(line: str) -> tuple[list[dict], list[str]]:
    fragments = _route_fragments(line)
    if not fragments:
        return [], []

    fragment_modes = [_mode_from_text(fragment) for fragment in fragments]
    whole_line_mode = None if any(fragment_modes) else _mode_from_text(line)
    legs = []
    pending = []
    previous_city = None

    for fragment, fragment_mode in zip(fragments, fragment_modes):
        cities, unknowns = _resolved_city_list(fragment, include_unknown=True)
        pending.extend(unknowns)
        if not cities:
            continue

        incoming_mode = fragment_mode or whole_line_mode
        if previous_city and cities[0] != previous_city:
            _append_leg(legs, previous_city, cities[0], incoming_mode)

        for index in range(len(cities) - 1):
            _append_leg(legs, cities[index], cities[index + 1], incoming_mode)

        previous_city = cities[-1]

    return legs, pending


def parse_route_text(text: str) -> dict:
    """Parse free-form itinerary text into map legs and route metadata."""
    legs = []
    pending = []
    arrival = None
    departure = None

    for raw_line in (text or "").splitlines():
        line = _strip_day_prefix(raw_line)
        if not line:
            continue

        parsed_arrival, parsed_departure, route_text = _extract_metadata(line)
        arrival = arrival or parsed_arrival
        departure = parsed_departure or departure

        if not route_text:
            continue

        line_legs, line_pending = _parse_route_line(route_text)
        legs.extend(line_legs)
        pending.extend(line_pending)

    return {
        "legs": legs,
        "pending": sorted(set(pending)),
        "arrival": arrival,
        "departure": departure,
    }